The well rate estimation challenge
Our team has worked on real production data from oil and gas operations for more than 10 years, on more than 15 assets globally and on more than 100 different use-cases. In our experience, the awareness of well flow rate quality and performance is low among production teams.
In a way, this is not so surprising. To be able to monitor and track such performance requires automation, integration and live data quality assessments of several data sources. Usually, the production engineers have a good idea of which of their well measurements are “trustworthy” and which are not. But an objective and quantifiable well rate measurement performance assessment is rarely available. And if you couple this lacking quantified assessment with the common industry view that the “well rate estimation problem” is solved, you have arrived at some sort of paradox. There are several solutions on the market, reporting error margins of about 5-15% (in theory), but no-one really knows how this plays out in real operations.
So with this as a backdrop, we decided to take a closer look at this question. We wanted to get a quantified sense of how the performance of MPFMs and VFMs are out there, and analyzed data from several of the assets we are connected to.
The performance range was found to be quite large, but interestingly most of the measurements achieve an error of 20% or less, in only about 50% of test points. The well flow rate estimation problem is not solved after all!
Solution - first commercial data-driven VFM offering
The well flow rate estimation problem is at the heart of true production optimization. Unless we know approximately what each well is producing at any given time, optimization in real-time becomes impossible.
Solution Seeker has spent several years developing a viable data-driven virtual flow metering solution, called NeuralCompass VFM. Using state-of-the-art machine learning methods and combining neural networks and multi-task learning with bayesian inference and online learning, we are proud to be the first company to offer the market a commercial data-driven VFM solution. Our research and development in this space is happening at great speed, and all our existing and potential clients can expect great advancements in this area. We have published several papers on the topic, with yet more in the pipeline. You can find our newest publications here:
NeuralCompass VFMs build upon our Squashy data mining algorithms, ensuring that high quality and relevant data are fed into the neural networks. In terms of accuracy and maintainability, our goal on each asset is to achieve less than 5% error on the well rate estimates with 90% less effort. Get in touch to hear more and see our latest results! Below you see an example of a high-fidelity NeuralCompass soft-tag together with the auto-detected stable Squashy intervals in gray.